An Improved Method for Detecting EDoS Attacks in the Cloud With Hyperparameter Optimization and Metaheuristic Algorithms

In recent years, cloud computing has gained significant popularity and adoption. As cloud infrastructure expands, vulnerabilities to attacks that exploit cloud services to drain paid resources also increase. A new form of attack, known as Economic Denial of Sustainability (EDoS) or Fraudulent Resource Consumption (FRC), is emerging as a major concern. This attack disrupts the cloud payment model by gradually escalating resource demand, leading to severe financial strain for both customers and service providers. EDoS attacks are classified as low-rate denial of service (LDoS) attacks, characterized by their covert nature and the challenges they pose to traditional detection methods. This paper introduces a methodology for detecting EDoS attacks by combining the Synthetic Minority Oversampling Technique (SMOTE) and the Edited Nearest Neighbor Rule (ENN) with Random Forest (RF) and XGBoost (XGB) classifiers. These classifiers are optimized using nature-inspired metaheuristics and conventional hyperparameter tuning methods, such as Random Search and Bayesian Search. Our comparative approach demonstrated improved performance, reduced computational costs, and decreased parameter search time. Notably, the application of the Bat Algorithm (BA) to optimize the XGBoost classifier showed superior results in identifying EDoS attacks, achieving higher accuracy (99.53%) and a lower error rate (0.46%) compared to other algorithms studied, including standard Random and Bayesian search methods.